一种基于双模控制的主动队列管理新算法  

A Novel Active Queue Management Algorithm Based on Double Model Control

作　　者：汪浩[1] 马学韬[1] 田作华[1] 

机构地区：[1]上海交通大学自动化系,上海200240

出　　处：《计算机仿真》2009年第8期112-115,127,共5页Computer Simulation

摘　　要：随机指数标记算法(REM)是一种有效的主动队列管理算法,但由于TCP/IP网络流量模型呈现非线性特性,故而其控制效果不佳,存在队列稳定性差,对动态流量响应慢等问题。为了解决上述问题,提出了基于双模控制的主动队列管理算法(Fuzzy-REM)。算法采用分段控制策略,在瞬时队列偏差大于阈值时,采用模糊控制,反之采用REM控制,从而将模糊控制的快速响应和REM稳态性能好的优点结合起来。NS2中的仿真实验表明,相对于REM算法,Fuzzy-REM提高了队列稳定性,加快了收敛速度,增强了算法对网络环境变化的适应性。Random exponential marking （REM） is an effective algorithm for active queue management （AQM）, but suffers from some problems such as big queue jitter and sluggish response to dynamic traffic due to the nonlinear property of TCP/IP network. To solve the above mentioned problems, a novel AQM algorithm based on double model control called Fuzzy - REM is proposed, which synthesizes the advantages of both fuzzy and REM controlling modes by using the segment control strategy. When the error of queue length is larger than the threshold, fuzzy controller is used to achieve high convergence speed, otherwise REM is used to get stable queue. The simulation experiments in NS2 demonstrate that Fuzzy - REM can achieve higher convergence speed, smaller queue jitter, and the adaptability to variable network conditions is enhanced effectively.

关 键 词：网络拥塞控制 主动队列管理 随机指数标记 模糊控制 双模控制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]